Posts
Comments
We ended up focusing on some other things in the past quarter, but we are about to plan the coming quarter, so we might end up prioritizing this again.
Adding yourself to the map is the first thing I would do, and then most likely I would see whether there have been any meetups historically anywhere close to your area. It appears there is a meetup in 10 days in Copenhagen, which seems pretty close to Malmo.
https://lesswrong.com/events/A7LAzFXD79pgFtsdA/cph-meetup-10-10-19
At that meetup you might also be able to figure out whether there are any people directly from your area interested in a meetup.
On the backend, we implemented a revision system for posts that allows us to mark edits to posts as "major revisions" which marks all the comments on that post as outdated, and also allows users to view past revisions of the post. As an example, see this post.
I hope we can soon get around to building some more user-facing UI for this, which allows users to edit a post while marking the revision as major, probably causing it to reappear on the frontpage.
We already implemented this!
When we set up crossposting we can set a flag on whether to have the canonical URL point towards its original source (this doesn't always make sense, for example for things like the AI Alignment Newsletter), but if you want to automatically crosspost while preserving the canonical URL we can set that up for you.
It's totally possible. In the post-editor, just select some empty space and press the "image" button in the toolbar. Or use markdown syntax.
Happy to edit the above post to have images in the post, as opposed to just links to images.
(I promoted this to frontpage, though I usually leave posts like this on personal blog, because it does actually cover a bunch of general lessons and I think is actually of pretty broad interest)
Yeah, I basically want this to replace the ToC as you scroll past the post and into the comment section.
This is correct.
We did not port over all of your old notifications, and from looking through Document's posts, it seems like they shouldn't have received any comment notifications since the relaunch, so my guess is they are looking for all the historical notifications, which I will sadly have to admit are no longer accessible in the same way (though importing them is definitely a possibility if enough people care, it just wasn't on our list of priorities and we've never gotten any request for them so far).
Yeah, this and a bunch of other reasons caused us to mostly make the call against link-coloring. We are currently going with annotations, though I would want to make them a lot smaller than they currently are on gwern.net (I think they are fine on gwern.net for the kind of content that you produced, but would be too distracting for LW content).
Indeed, thanks! Added them to the list above
We don't really control the LW Wiki, so not sure how feasible this is. It would probably require a completely separate codepath, and I don't know what the subscription infrastructure of the wiki has. I will look into it, but can't promise anything as of yet.
Some opportunities that I know about:
- The Long Term Future Fund is I think currently the most active granter in the space for independent researchers
- MIRI has in the past funded some independent researchers, and some programs that try to facilitate research (like funding to MIRIx workshops, or the AI Safety camp, or related projects)
- Jaan Tallinn has some budget set aside to fund people doing interesting research, I think partially administered via Alex Zhu. Happy to put anyone interested into contact.
- There exist a variety of small donors who I think are open to this. Matt Wage (also on the LTFF) is often open to things like this.
- The Open Philanthropy Project has (had?) a scholarship program, here are some examples of people they funded, I do not know whether there is any way to apply to it:
- https://www.openphilanthropy.org/focus/global-catastrophic-risks/scholarship-support-2019
- https://www.openphilanthropy.org/focus/global-catastrophic-risks/potential-risks-advanced-artificial-intelligence/ai-scholarships-2018
- The Open Philanthropy Project also has an AI Policy careers scholarship program:
- The EA Foundation has given grants for AI Alignment Research:
- Being a student at CHAI is one way of getting funding. They might also be open to more independent research arrangements, but not sure:
- The FHI research scholars program includes people who work on technical AI Alignment:
- BERI has historically done some grantmaking in this space, though that will likely now be done via the Survival and Flourishing Fund. For that, applications are open for 4 more days, though require some kind of existing institutional affiliation.
- I can imagine that Tyler Cowen's Emergent Ventures is open to funding things in this space
Will edit this if I come up with anything else, or if someone else points out something that I am missing.
Edit: New additions by Issa Rice:
- Paul Christiano has done some independent funding in this space (and maybe Open AI is also open to that, but I don't know)
- FLI has made grants, though I don't know about the current state of their grant program
- EA Grants also has given some grants in this space
I felt a bit uncertain about doing one every month, and was planning to start another one in October. Depending on how that one goes we might go with a monthly schedule, or maybe every two months is the right way to go.
I think you are overestimating the degree to which there was "no real tradeoff". I bet with you that many people felt tempted to enter the launch codes, for reasons that Said mentioned, but also as a resource to be sold (like Jeff Kauffman considered).
Like, it's obvious that if you didn't do the thing with the codes, and just had a button available to every LessWrong user, someone would have pressed it "just for the lulz", which isn't a very good reason to press the button, but as you scale the number of people who are trying to balance the pros and cons, someone will think that the lulz are more important than not pressing the button (and I think "for the lulz" can be pretty validly described as someone doing something for the public good).
I was quite uncertain whether anyone would press the button over the course of the day, and don't think that uncertainty was unjustified. And I would be surprised if you were confident at 90%+ that no one would press the button over the course of the day.
I do think that given how it worked out this time, adding a more concrete ritualized incentive seems maybe good to me, like giving $500 to a charity chosen by whoever presses the button.
I also don't think the constraint of "what matters are only considerations about the public good" is a good constraints for a tradition like this, because in reality obviously many people will not only think about the public good, which actually significantly worsens the problems that the unilateralist curse is talking about, and it seems bad to not incorporate those into this tradition (and the UC is the best abstraction we currently have for talking about this reference class).
Cool, will transfer ownership and make this a draft
Mechanics for donating to LessWrong:
LessWrong is sponsored by CFAR, so if you just make a donation to CFAR and earmark it for the LessWrong project it will show up in our bank accounts. For anything larger than a few hundred dollars it's probably best to send me an email at oliver.habryka@gmail.com and then figure out the details from there.
For smaller donations just using the CFAR PayPal, with a note saying it's for the LessWrong project, and also sending me a small email to make sure it ends up in the right accounts is likely the best thing to do.
Hmm, sure. Let's space them out at like one a week.
Which part of the two statements? That destroying trust is really bad, or that the case hasn't been made?
Nice, I activated them and if it overall seems good, we will do the same for LessWrong.
Jeff does conveniently have a blogpost on this: https://www.jefftk.com/p/what-should-counterfactual-donation-mean
We had some original plans of coordinating with the EA Forum people on this, but didn't end up having enough time available to make all of that happen. Agree that the ideal reenactment scenario would include two forums (though with mutually assured destruction in the later parts of the cold war, the outcome is ultimately the same).
My sense is that it's very unlikely to be worth it at anything below $10k, and I might be a bit tempted at around $50k, though still quite hesitant. I agree that at $1M it's very likely worth it.
The site will go down for a full 24 hours after the button was pressed and correct launch codes entered (not that that is the most important aspect of this situation, but I figured I would clarify anyways)
Yep, you have to activate the map to see it. Just turned out to be the most convenient way of implementing it, and also worked well aesthetically.
Contrarian opinion: The two terms are not unrelated and have a surprising amount of actual shared structure.
Everyone has access to the button, but only 125 users were sent launch codes (after you press the button the launch-codes panel appears).
Admins can mark a new version of a post as a "major revision" which allows you to see the history of all major revisions on the post page. For normal posts there isn't currently a way to see the version history, though it's technically accessible through the API and if you have any specific posts, I would be happy to make the version history accessible for that one.
For your own posts, we have some plans for making the version history accessible in the editor, as part of the editor rework.
(note, this is now fixed)
Ah, good to know. Sorry for the hassle!
My biggest concern with this song has always been the refrain, which I failed to make any sense of, and so I kind of feel like the song doesn't resonate with me:
But I am willing to sacrifice
Something we don't have for something we won't have
What is the thing that we don't have, and how can you sacrifice something you don't have? It seems like it should just straightforwardly say "something I have", but that doesn't fit into the meter.
And then also the fact that my timelines are relatively short, so it's actually not really the case that I am sacrificing it for something we won't have, since I assign decent probability to the current generation living forever. Though that part feels more fine than the weird negation in the refrain.
Welcome!
Looking forward to seeing more of your ideas written up, a few of them have been quite valuable to me over the years, and it seems great to make them available to a wider audience.
Alas, that is very sad. Let me know what you find, so I can make sure this doesn't happen again.
Promoted to curated: I really like this sequence, and I think this post might actually be one of the best in the sequence, due to the degree to which it tries really hard to create concrete technical models of how the human mind works (which I think also makes it a bit less accessible than some of the other posts in the sequence).
I am generally a fan of minimalism, and having a focused reading experience. I like being able to read a post, and just focus on the content, without lots of other stuff that is trying to get my attention, until I am done with it.
This is mostly based on my models of intellectual effort, where I think distraction-rich environment make putting real effort into something a lot harder.
Ah, I see. Sorry for the confusion. I usually say "Mod note" these days.
I am hesitant about sidebars in general, but I agree that having recommendations below the post at least is a good idea. I would also experiment with putting them in the ToC navigation, though you don't really have space for more than 2-3 related posts there.
I don't remember the details, but I think it was more that the discussion was generally low-quality, not because of the political content in particular.
This has since been changed. Link posts are now identified by the initial paragraph saying "This is a linkpost for X". No small hidden link icons.
What is the purpose of karma?
LessWrong has a karma system, mostly based off of Reddit's karma system, with some improvements and tweaks to it. I've thought a lot about more improvements to it, but one roadblock that I always run into when trying to improve the karma system, is that it actually serves a lot of different uses, and changing it in one way often means completely destroying its ability to function in a different way. Let me try to summarize what I think the different purposes of the karma system are:
Helping users filter content
The most obvious purpose of the karma system is to determine how long a post is displayed on the frontpage, and how much visibility it should get.
Being a social reward for good content
This aspect of the karma system comes out more when thinking about Facebook "likes". Often when I upvote a post, it is more of a public signal that I value something, with the goal that the author will feel rewarded for putting their effort into writing the relevant content.
Creating common-knowledge about what is good and bad
This aspect of the karma system comes out the most when dealing with debates, though it's present in basically any karma-related interaction. The fact that the karma of a post is visible to everyone, helps people establish common knowledge of what the community considers to be broadly good or broadly bad. Seeing a an insult downvoted, does more than just filter it out of people's feeds, it also makes it so that anyone who stumbles accross it learns something about the norms of the community.
Being a low-effort way of engaging with the site
On lesswrong, Reddit and Facebook, karma is often the simplest action you can take on the site. This means its usually key for a karma system like that to be extremely simple, and not require complicated decisions, since that would break the basic engagement loop with the site.
Problems with alternative karma systems
Here are some of the most common alternatives to our current karma system, and how they perform on the above dimensions:
Eigenkarma as weighted by a set of core users
The basic idea here is that you try to signal-boost a small set of trusted users, by giving people voting power that is downstream from the initially defined set of users.
There are some problems with this. The first one is whether to assign any voting power to new users. If you don't you remove a large part of the value of having a low-effort way of engaging with your site.
It also forces you to separate the points that you get on your content, from your total karma score, from your "karma-trust score" which introduces some complexity into the system. It also makes it so that increases in the points of your content, no longer neatly correspond to voting events, because the underlying reputation graph is constantly shifting and changing, making the social reward signal a lot weaker.
In exchange for this, you likely get a system that is better at filtering content, and probably has better judgement about what should be made common-knowledge or not.
Prediction-based system
I was talking with Scott Garrabrant today, who was excited about a prediction-based karma system. The basic idea is to just have a system that tries to do its best to predict what rating you are likely to give to a post, based on your voting record, the post, and other people's votes.
In some sense this is what Youtube and Facebook are doing in their systems, though he was unhappy with the transparency of what they were doing.
The biggest sacrifice I see in creating this system, is the loss in the ability to create common knowledge, since now all votes are ultimately private, and the ability for karma to establish social norms, or just common knowledge about foundational facts that the community is built around, is greatly diminished.
I also think it diminishes the degree to which votes can serve as a social reward signal, since there is no obvious thing to inform the user of when their content got votes on. No number that went up or down, just a few thousand weights in some distant predictive matrix, or neural net.
Augmenting experts
A similar formulation to the eigenkarma system is the idea of trying to augment experts, by rewarding users in proportion to how successful they are at predicting how a trusted expert would vote, and then using that predicted expert's vote as the reward signal. Periodically, you do query the trusted expert, and use that to calibrate and train the users who are trying to predict the expert.
This still allows you to build common-knowledge, and allows you to have effective reward signals ("simulated Eliezer upvoted your comment"), but does run into problems when it comes to being a low-effort way of engaging with the site. The operation of "what would person X think about this comment" is a much more difficult one than "did I like this comment?", and as such might deter a large number of users from using your site.
Yeah, I also found that yesterday. Will fix that today.
Done!
Promoted to curated: I like the whole sequence, and think it's elaborating and extending on existing writing on the site in a really valuable way, and this post seemed like the best place to start, as well as one that is just robustly good in a variety of ways.
I particularly appreciate the concreteness of the whole sequence, and the extensive use of examples and small exercise-like things.
Agree that they have advantages. I personally much prefer email notifications, because it allows me to centralize all my notification systems, and because I generally prefer to process things in batches, but I can see the case for app notifications.
I will look into this more. Will let you know what I find.
Yeah, this was originally known as "Anti-Kibitzer" on the old LessWrong. It isn't something we prioritized, but I think greaterwrong has an implementation of it. Though it would also be pretty easy to create a stylish script for it (this hides it on the frontpage, and makes the color white on the post-page, requiring you to select the text to see the score):
https://userstyles.org/styles/175379/lesswrong-anti-kibitzer
Was definitely intentional, though I think mostly for relatively bad reasons that I should get around to fixing. I could explain, but I think the original intention was mostly just wrong, so let's treat it as a bug for now.
Ah, I see. Then yes, I agree with you that better Inbox management for notifications in general seems quite valuable. As I mentioned, we have an upcoming rework of the notification system in general, so after that is done, I will look into doing that.
Also agree with backlinks. We are about to roll out hover link-previews, which is a related feature, so I think we can do that soon.
Re the last linked feature: We also have some drafts and plans for a react-system, that I am pretty excited about implementing, but held off on for the last three months because we decided to focus on some stuff that would conflict with that, but I expect us to get around to that soon.
I added an "archive" button to the PM system a few months ago, which I was hoping would serve some of the function you were requesting. Adding some kind of tags or flags to it also seems like a reasonable improvement, though not that many users use the LessWrong PM system, so I am kind of hesitant to optimize that part of the site super much.
Re development speed: I think it's mostly that we've been prioritizing other stuff. I.e. I've been revamping a bunch of the map infrastructure and meetup infrastructure recently, which I think is something you are not super interested in, but I think is pretty important for other reasons. Most of our feature prioritization is a result of some long-term visions, and user interviews we make. The currently two most requested features we are working on is adding Google Docs-like sharing functionality to posts, which is a relatively major project, as well as a revamp of the notification system that also supports more granular email notifications, neither of which are mentioned in this thread, but are definitely the most requested features via all the other channels we have for people making feature requests.
(And actually, checking the rest of this thread, there isn't a single feature request on it that I feel like I want to prioritize, that isn't already implemented, besides the two comments you made today. Is there anything else in this thread that you think we should prioritize?)
We've been working on a revamp of the notification system, which is at least related to this. Would email notifications with a daily digest (and settings to be informed of certain events individually and immediately) be sufficient for a lot of these, or is there something specific about the push-notifications from the app that would help with this?
I've been thinking about an app for a while. For notifications, I actually think just making LW a full progressive web app and using the notifications API of mobile browsers seems like the best choice to me, but not confident. I don't think anyone on the team has super much experience with native App development, so building a full native app would definitely be at least a bit of a learning effort.
At the very least "The New Atlantis", a fictional utopian novel he wrote, was quite influential, at least in that it's usually cited as one of the primary inspirations for the founding of the royal society:
Lol, thanks. The next meetup on my list was "Brisbane, Australia" so I mixed those two together.